Last month saw the introduction of Yahoo Pipes, a tool for graphically creating and sharing customized content feeds from a variety of sources. Initial reaction from the blogosphere was largely positive, but I found it more difficult than I expected to create my first pipe.

Pipes’ drag and drop interface reminds me very much of STELLA, which has been around for more than twenty years and is still used in teaching ecological modeling. Changing how the modules connect doesn’t work the way I expected – I keep grabbing at the wrong parts. My layouts always look like spaghetti until I click on the Layout button.

What really slowed me down was that I was convinced that the heart of the matter was the Content Analyzer module, which is not well described. There’s no way to look under the hood. It applies some hidden algorithm to extract salient keywords (output as a y:content_analyzer string) that you can then pipe to a filter. However in the examples I tried, there wasn’t much added value in this content_analyzer. I still have to tweak filters to suit the categorizing, tagging, and writing styles of the different sources.

On the positive side, after the initial learning curve, I created two simple pipes for a couple of my projects: one finds butterfly and moth photos on Flickr given a zipcode provided by a user, the other combs feeds from a variety of news sources for stories on wildlife and biodiversity . Of course I’ve published these so others can subscribe to them or clone them. The next step will be to embed them in my project pages so they’ll run for our visitors.

This system is already powerful. If Yahoo! adds some of the features and modules suggested by Shelley Powers and folks on the discussion boards it will be even more useful. The Location Input module should allow latlongs or place names and allow a user to set a radius for “near.” I’d also suggest some way to incorporate semantics in a configurable Content Analyzer. For example, if I could ask for Lepidoptera, point to the appropriate ontology, and apply a little reasoning, I could avoid laborious filters currently needed to combine results for “skipper”, “moth”, and “butterfly” and reject those having to do with sailors and operas.

Categories: Uncategorized, , Related posts: • Emergent semantics Yahoo group;  • Environmental detection/protection.;  • Building a XMPP GTalk agent;   Comments: